Srikrishnan V Email and Phone Number

Working on Computer Vision algorithms for ADAS domain.

I worked on imaging problems in Image Guided Therapy and Ultrasound imaging.

In this role, I led a mid-size team consisting of algorithm and software engineers. We work on diagnostic medical images spanning across modalities like CT/PET/MRI. We tackle segmentation, classification, landmark and object(organ) localisation problems using a judicious mix of Deep Learning and other methods.

I worked in the X-Ray or Digital Radiography division at GE Healthcare. In this role, I led the effort on camera based workflow improvements. Beginning with POCs, I was able to take this to a commercialisation with rapid iterations of algorithm and software development based on feedback. This was launched as a commercial offering with Live video streaming and advanced features like Position Assist and Technic Assist. I have also filed IPs in this area. The link of the commercial offering is below:

I worked both as an individual contributor and also (technically) managed a small team of about 5 to 6 engineers. The problems we were solving was Brain volumetry, brain metastasis segmentation, cardiac segmentation, perfusion analysis and knee cartilage segmentation. We employed a variety of methods: ranging from mutli-atlas registration, machine learning techniques kNN/EM, deconvolution and of course CNNs for cartilage segmentation. This was around the time CNNs was getting popular in medical. We also had a gala time building caffe on Windows with different compilers. We were able to publish our methods and results in conferences.However, at one point of time Samsung decided not to pursue MRI.

Primarily, I worked on algorithms for MRI image reconstruction and postprocessing applications. There was no commercialisation involved. We were able to apply the state of art de-noising algorithms like NLM, BM3D etc. onto MRI images. The challenge lay in having to optimise the algorithm parameters for different acquisitions and anatomies automatically. We were able to present good quality de-noised images to the clinicians.

This was my first experience in medical imaging. I was fortunate enough to get a chance to work on an exciting PET/CT problem involving image quality improvement using image registration. Although a lot of work had been done in this project, I was able to contribute a simple idea which was critical for the commercial launch. IPs were also filed.

For PhD, I worked on tracking and segmentation of objects in video and images. My primary work was in proposing a tangential curve evolution term, which stabilises parametric curve evolution. The C code is available; it uses OpenCV but is not highly dependent on it. Apart from curve evolution, the thesis has minor contributions in the planar and point based tracking applications.

I spent an graduate engineer trainee for one year trying to understand how UPS and the organisation work. I left to join a Master's programme at I.I.T Bombay.